Risk assessment method and system based on multi-modal data, terminal and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a risk assessment method and system based on multi-modal data, a terminal and a storage medium, and the method comprises the steps: obtaining the multi-modal data of a patient, carrying out the matching and calculation based on a preset VTE risk assessment rule according to the multi-modal data, and obtaining a basic risk level and a basic prevention strategy; obtaining dynamic time sequence data of a patient, adjusting the basic risk level according to the dynamic time sequence data, generating a plurality of dynamic risk weights, and performing fusion calculation on all the dynamic risk weights to obtain a scoring result; and obtaining an early warning trigger condition, determining a target intervention strategy in the basic prevention strategies according to the early warning trigger condition and the scoring result, and executing the target intervention strategy. According to the method, the VTE risk of the patient is iteratively optimized through a dynamic weight fusion mechanism, and the accuracy of risk assessment and intervention strategies is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to a risk assessment method and system based on multi-modal data, a terminal and a computer readable storage medium. BACKGROUND

[0002] At present, traditional VTE (Venous Thromboembolism) assessment relies on manual filling of Caprini (Caprini Risk Assessment Model) score table, which takes 10-15 minutes per case, and is prone to missed evaluation due to work negligence; the static assessment mode is only carried out at fixed nodes such as admission and postoperative, and cannot track the changes of the disease in real time (such as dynamic increase of D-dimer after operation); the preventive measures depend on experience decision, and there is a problem of insufficient balance between bleeding risk and thrombosis risk (such as inaccurate identification of contraindications for anticoagulation).

[0003] The traditional VTE assessment method lacks a dynamic monitoring mechanism, and it is difficult to capture the sudden change of acute thrombosis risk, and the multi-role cooperation depends on information silos, so it is difficult to obtain accurate information in time, which further leads to misjudgment, and thus in the case of traditional VTE assessment, it is impossible to track the changes of the disease in real time and make judgments in time due to the dependence on manual static filling, which has become a problem to be solved at present.

[0004] Therefore, the prior art still needs to be improved and developed. SUMMARY

[0005] The main purpose of the present application is to provide a risk assessment method and system based on multi-modal data, a terminal and a computer readable storage medium, which aims to solve the problem that the traditional VTE assessment method in the prior art cannot track the changes of the disease in real time and make judgments in time.

[0006] To achieve the above-mentioned purpose, the present application provides a risk assessment method based on multi-modal data, which comprises the following steps: Obtaining multi-modal data of a patient, matching and calculating the multi-modal data based on a preset VTE risk assessment rule to obtain a basic risk level and a basic prevention strategy; Obtaining dynamic time series data of the patient, adjusting the basic risk level according to the dynamic time series data to generate a plurality of dynamic risk weights, and fusing and calculating all the dynamic risk weights to obtain a score result; Obtaining a warning trigger condition, determining a target intervention strategy in the basic prevention strategy according to the warning trigger condition and the score result, and executing the target intervention strategy.

[0007] Optionally, the risk assessment method based on multi-modal data, wherein the multi-modal data comprises electronic medical record text and ultrasound report. The multi-modal data of the patient is obtained, and the multi-modal data is matched and calculated based on a preset VTE risk assessment rule to obtain a basic risk level and a basic prevention strategy, specifically comprising: The electronic medical record text and the ultrasound report of the patient are obtained, and the electronic medical record text and the ultrasound report are matched based on a preset VTE risk assessment rule to obtain a plurality of evaluation values. The basic risk level is calculated according to all the evaluation values, and the basic prevention strategy is determined according to the basic risk level. The basic risk level comprises any one of a low basic level, a sub-low basic level, a medium basic level, and a high basic level.

[0008] Optionally, the risk assessment method based on multi-modal data, wherein the dynamic time series data comprises operation type, bleeding amount, D-dimer time series data, and vital sign data. The dynamic time series data of the patient is obtained, the basic risk level is adjusted according to the dynamic time series data, a plurality of dynamic risk weights are generated, and all the dynamic risk weights are fused and calculated to obtain a score result, specifically comprising: The operation type, the bleeding amount, the D-dimer time series data, and the vital sign of the patient are obtained. The basic risk level is adjusted according to the operation type, the bleeding amount, the D-dimer time series data, and the vital sign to obtain a plurality of risk weight values, and all the risk weight values are fused and calculated to obtain a score result.

[0009] Optionally, the risk assessment method based on multi-modal data, wherein the basic risk level is adjusted according to the operation type, the bleeding amount, the D-dimer time series data, and the vital sign to obtain a plurality of risk weight values, and all the risk weight values are fused and calculated to obtain a score result, specifically comprising: The operation type, the bleeding amount, the D-dimer time series data, and the vital sign are input into a deep learning model for analysis to obtain an analysis result. The low basic level, the sub-low basic level, the medium basic level, and the high basic level are adjusted according to the analysis result to obtain a plurality of risk weight values, and all the risk weight values are fused and calculated to obtain a score result.

[0010] Optionally, in the risk assessment method based on multimodal data, the analysis results include any one of low-risk, medium-risk, and high-risk results; The process involves inputting the surgical type, blood loss, D-dimer time-series data, and vital signs into a deep learning model for analysis to obtain the analysis results, specifically including: The surgical type, the amount of blood loss, the D-dimer time series data, and the vital signs are divided to obtain a training set and a test set. Based on the initial deep learning model, the LSTM network of the initial deep learning model is trained according to the training set to obtain the deep learning model. The test set is input into a deep learning model for analysis to obtain any one of the following results: low risk, medium risk, or high risk.

[0011] Optionally, in the risk assessment method based on multimodal data, the early warning triggering conditions include core triggering conditions, high-risk factor triggering conditions, and dynamic change triggering conditions; The process of obtaining early warning trigger conditions, determining a target intervention strategy within the basic prevention strategy based on the early warning trigger conditions and the scoring results, and executing the target intervention strategy specifically includes: Obtain the current stage of diagnosis and treatment, along with the core triggering conditions, the high-risk factor triggering conditions, and the dynamic change triggering conditions; Based on the core triggering conditions, the high-risk factor triggering conditions, the dynamic change triggering conditions, and the scoring results, a target intervention strategy is determined in the basic prevention strategy, and the target intervention strategy is executed.

[0012] Optionally, in the risk assessment method based on multimodal data, the target intervention strategy includes either an anticoagulation strategy or an avoidance strategy. The step of determining a target intervention strategy within the basic prevention strategy based on the core triggering conditions, the high-risk factor triggering conditions, the dynamic change triggering conditions, and the scoring results, and then executing the target intervention strategy, specifically includes: If the analysis result is a medium-risk or high-risk result, then the anticoagulation strategy is determined in the basic prevention strategy based on the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition, and the scoring result. If the analysis result is not a medium-risk or high-risk result, then the avoidance strategy is determined in the basic prevention strategy based on the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition, and the scoring result.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a risk assessment system based on multimodal data, wherein the risk assessment system based on multimodal data includes: The multimodal data acquisition module is used to acquire the patient's multimodal data, and based on the preset VTE risk assessment rules, it matches and calculates the multimodal data to obtain the basic risk level and basic prevention strategy; The risk assessment module is used to acquire the patient's dynamic time-series data, adjust the basic risk level based on the dynamic time-series data, generate multiple dynamic risk weights, and fuse all the dynamic risk weights to obtain a score result. The intervention strategy execution module is used to acquire the early warning triggering conditions, determine the target intervention strategy in the basic prevention strategy based on the early warning triggering conditions and the scoring results, and execute the target intervention strategy.

[0014] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a risk assessment program based on multimodal data, and when the risk assessment program based on multimodal data is executed by a processor, it implements the steps of the risk assessment method based on multimodal data as described above.

[0015] In this invention, multimodal data of the patient is acquired, and based on preset VTE risk assessment rules, the multimodal data is matched and calculated to obtain a basic risk level and a basic prevention strategy. Dynamic time-series data of the patient is acquired, and the basic risk level is adjusted based on the dynamic time-series data to generate multiple dynamic risk weights. All dynamic risk weights are then fused and calculated to obtain a score. Early warning triggering conditions are acquired, and a target intervention strategy is determined within the basic prevention strategy based on the early warning triggering conditions and the score, and the target intervention strategy is executed. This invention iteratively optimizes the patient's VTE risk through a dynamic weight fusion mechanism, significantly improving the accuracy of risk assessment and intervention strategies. Attached Figure Description

[0016] Figure 1 This is a flowchart of a preferred embodiment of the risk assessment method based on multimodal data of the present invention; Figure 2 This is a flowchart of S10 of a preferred embodiment of the risk assessment method based on multimodal data of the present invention; Figure 3 This is a flowchart of S20 of a preferred embodiment of the risk assessment method based on multimodal data of the present invention; Figure 4 This is a flowchart of S22 of a preferred embodiment of the risk assessment method based on multimodal data of the present invention; Figure 5This is a flowchart of S222 of a preferred embodiment of the risk assessment method based on multimodal data of the present invention; Figure 6 This is a flowchart of the scoring results of a preferred embodiment of the risk assessment method based on multimodal data of the present invention; Figure 7 This is a flowchart of S30 of a preferred embodiment of the risk assessment method based on multimodal data of the present invention; Figure 8 This is a flowchart of S32 of a preferred embodiment of the risk assessment method based on multimodal data of the present invention; Figure 9 This is a flowchart illustrating the generation of a personalized prevention scheme, a preferred embodiment of the risk assessment method based on multimodal data of the present invention. Figure 10 This is a structural diagram of a preferred embodiment of the risk assessment system based on multimodal data of the present invention; Figure 11 This is a structural diagram of a preferred embodiment of the terminal of the device of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0018] Traditional VTE assessment methods lack dynamic monitoring mechanisms, making it difficult to capture acute-phase thrombotic risk mutations. Multi-role collaboration relies on information silos, hindering the timely acquisition of accurate information and leading to misjudgments. Consequently, traditional VTE assessments, due to their reliance on manual, static data entry, cannot track changes in the patient's condition in real time and make timely judgments. Therefore, a risk assessment method based on multimodal data is needed. This method should iteratively optimize the patient's VTE risk through a dynamic weight fusion mechanism, significantly improving the accuracy of risk assessment and intervention strategies, and enabling real-time tracking of changes in the patient's condition and timely judgments.

[0019] The risk assessment method based on multimodal data described in the preferred embodiment of the present invention, such as... Figure 1 As shown, the risk assessment method based on multimodal data includes the following steps: Step S10: Obtain the patient's multimodal data, and perform matching and calculation based on the multimodal data according to the preset VTE risk assessment rules to obtain the basic risk level and basic prevention strategy.

[0020] like Figure 2 As shown, step S10 includes: Step S11: Obtain the patient's electronic medical record text and ultrasound report, and match the electronic medical record text and ultrasound report according to the preset VTE risk assessment rules to obtain multiple assessment values; Step S12: Calculate the basic risk level based on all the assessment values, and determine the basic prevention strategy based on the basic risk level.

[0021] Specifically, the patient's electronic medical record text and ultrasound report (electronic medical record text (past medical history, comorbidities), ultrasound report (lower extremity venous ultrasound)) are acquired. Based on preset VTE risk assessment rules, the electronic medical record text and ultrasound report are matched to obtain multiple assessment values ​​(assessment results: 0 points for low-level basic risk, 1-2 points for sub-low-level basic risk, 3-4 points for intermediate-level basic risk, ≥5 points for high-level basic risk). The basic risk level is calculated based on all the assessment values, and a basic prevention strategy is determined based on the basic risk level (low-level or sub-low-level basic risk: health education + basic prevention; intermediate-level basic risk: health education + basic prevention + physical prevention; high-level basic risk: health education + basic prevention + physical prevention + drug prevention). The multimodal data includes the electronic medical record text and ultrasound report, wherein the basic risk level includes any one of the low-level, sub-low-level, intermediate-level, and high-level basic risk levels.

[0022] In this embodiment, health education includes: encouraging adequate water intake to avoid blood concentration and regular monitoring of coagulation function and complete blood count; dietary guidance: low-fat, bland diet, avoiding spicy, stimulating, and fatty foods; advising patients to improve their lifestyle, such as quitting smoking and alcohol, and controlling blood sugar and lipids; informing patients to promptly notify medical staff if they experience symptoms such as limb swelling, pain, chest pain, or difficulty breathing; maintaining a positive mood; basic prevention: guiding patients to perform active ankle pump exercises, 20-30 times / hour; encouraging patients to get out of bed early; using pillows or a rocking bed to elevate the affected limb 20-30 cm above heart level; providing daily guidance and assistance for active and passive exercises; massaging the calf muscles with a circumferential compression technique; avoiding placing hard pillows under the knees and excessive hip flexion; checking Homans' sign every shift; mechanical prevention: administering intermittent pneumatic compression therapy as prescribed; using a plantar venous pump as prescribed; wearing graded compression stockings as prescribed; and drug prevention: observing for signs of bleeding; regularly monitoring coagulation function and complete blood count; and administering antithrombotic drugs as prescribed.

[0023] Step S20: Obtain the patient's dynamic time-series data, adjust the basic risk level based on the dynamic time-series data, generate multiple dynamic risk weights, and fuse all the dynamic risk weights to obtain the scoring result.

[0024] like Figure 3 As shown, step S20 includes: Step S21: Obtain the patient's surgical type, blood loss, D-dimer time-series data, and vital signs; Step S22: Adjust the basic risk level according to the surgical type, the amount of bleeding, the D-dimer time series data and the vital signs to obtain multiple risk weight values, and fuse all the risk weight values ​​to obtain the scoring result.

[0025] Specifically, the following information is obtained: the patient's surgical type, blood loss, D-dimer time-series data (D-dimer time-series data refers to a series of D-dimer measurements collected at different time points for the same individual; these data are not merely isolated values ​​but contain a trend that changes over time; D-dimer time-series data itself is a unimodal data type (i.e., time-series numerical data), but in practical applications, it is always integrated into multimodal data and used in conjunction with other modalities such as text and images for comprehensive patient assessment, disease diagnosis, treatment effect monitoring, and prognosis judgment), and vital signs (vital signs are the most core and fundamental indicators for measuring basic human physiological functions. When constructing time-series features for clinical early warning (such as VTE risk warning), these...) It is one of the most important and continuous data sources. The four traditional vital signs include: body temperature, pulse (heart rate), respiratory rate, and blood pressure. The basic risk level is adjusted according to the surgical type, blood loss, D-dimer time series data, and vital signs to obtain multiple risk weight values. All the risk weight values ​​are merged and calculated to obtain the score result (e.g., age ≥75 years: 3 points, acute myocardial infarction (within 1 month): 1 point, congestive heart failure: 1 point, bed rest >72h: 2 points, total score: 3+1+1+2=7 points). The dynamic time series data includes surgical type, blood loss, D-dimer time series data, and vital sign data.

[0026] like Figure 4 As shown, step S22 includes: Step S221: Input the surgical type, the amount of blood loss, the D-dimer time series data, and the vital signs into a deep learning model for analysis to obtain the analysis results; Step S222: Adjust any one of the low-level basic level, the second-lower-level basic level, the intermediate-level basic level, and the high-level basic level according to the analysis results to obtain multiple risk weight values. Combine all risk weight values ​​to obtain the scoring result.

[0027] Specifically, the surgical type, blood loss, D-dimer time-series data, and vital signs are input into a deep learning model for analysis to obtain analysis results (deep learning model: LSTM network processes D-dimer time-series data to obtain analysis results, Transformer model analyzes electronic medical record text). Based on the analysis results, any one of the low-level basic grade, the second-lower-level basic grade, the intermediate-level basic grade, and the high-level basic grade is adjusted to obtain multiple risk weight values. All risk weight values ​​are then fused and calculated to obtain the scoring result.

[0028] like Figure 5 As shown, step S222 includes: Step S2221: Divide the surgical type, the amount of bleeding, the D-dimer time series data and the vital signs to obtain a training set and a test set. Based on the initial deep learning model, train the LSTM network of the initial deep learning model according to the training set to obtain a deep learning model. Step S2222: Input the test set into the deep learning model for analysis to obtain any one of the following: low-risk result, medium-risk result, and high-risk result.

[0029] Specifically, the surgical type, the amount of bleeding, the D-dimer time-series data, and the vital signs are divided to obtain a training set and a test set. Based on the initial deep learning model, the LSTM network of the initial deep learning model is trained according to the training set to obtain a deep learning model. The test set is input into the deep learning model for analysis to obtain any one of low-risk, medium-risk, and high-risk results. The analysis result includes any one of the following: low-risk, medium-risk, and high-risk results.

[0030] For example, the surgical type, blood loss, D-dimer time-series data, and vital signs are all crucial in two stages: 1. Model training stage, and 2. Model prediction (application) stage. Therefore, electronic medical record text serves both as the necessary training material (training set) for building an intelligent VTE system and as the raw material (predictive input) for real-time analysis in actual operation. The entire process relies on natural language processing technology to bridge the gap between unstructured text and structured risk factors.

[0031] In this embodiment, as Figure 6As shown, the system automatically acquires patient data based on the hospital information system (HIS / EMR / LIS), calculates and outputs specific scores using the Caprini scoring model, and the system's built-in judgment logic evaluates in real time whether the score is greater than or equal to 5. If this threshold is reached, an extremely high-risk warning is automatically triggered, and the warning action is executed through pop-up reminders, interface highlighting, or message sending. Finally, the system presents the warning information and evaluation results to the doctor, who then makes a diagnosis and treatment decision based on the clinical situation and records it in the system.

[0032] Step S30: Obtain the early warning triggering conditions, determine the target intervention strategy in the basic prevention strategy based on the early warning triggering conditions and the scoring results, and execute the target intervention strategy.

[0033] like Figure 7 As shown, step S30 includes: Step S31: Obtain the current diagnosis and treatment stage, the core triggering condition, the high-risk factor triggering condition, and the dynamic change triggering condition; Step S32: Determine the target intervention strategy in the basic prevention strategy based on the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition, and the scoring result, and execute the target intervention strategy.

[0034] Specifically, the early warning triggering conditions include core triggering conditions, high-risk factor triggering conditions, and dynamic change triggering conditions. The current diagnosis and treatment stage is obtained along with the core triggering conditions, high-risk factor triggering conditions, and dynamic change triggering conditions. Based on the core triggering conditions, high-risk factor triggering conditions, dynamic change triggering conditions, and the scoring results, a target intervention strategy is determined in the basic prevention strategy, and the target intervention strategy is executed. The early warning triggering conditions include core triggering conditions, high-risk factor triggering conditions, and dynamic change triggering conditions.

[0035] In this embodiment, the core triggering condition (based on standardized scoring) is the most common and systematic triggering method. The Caprini (Venous Thromboembolism Risk Assessment Model) score reaches the high / very high risk threshold. The triggering condition is that the Caprini score automatically calculated by the system is ≥ 5 points. The Caprini score is very detailed and includes dozens of risk factors (such as age, type of surgery, medical history, etc.). The higher the score, the greater the risk. 1-2 points: low risk; 3-4 points: intermediate risk; ≥ 5 points: high risk. For example, a 68-year-old patient (2 points) undergoes laparoscopic radical resection of lung cancer (2 points) due to lung cancer (3 points). His total Caprini score is 7 points. If the score is ≥ 5 points, the system will immediately trigger a high-risk warning. The Padua score reaches the high-risk threshold. The triggering condition is that the Padua score automatically calculated by the system is ≥ 4 points. Note: The Padua scale contains 11 risk factors and is more suitable for non-surgical medical patients. A score <4 indicates low risk; a score ≥4 indicates high risk. For example, a 75-year-old (3 points) heart failure patient (1 point) who is admitted to the hospital due to an acute infection and requires absolute bed rest (3 points) has a total Padua score of 7. A score ≥4 triggers a high-risk warning.

[0036] As an example, the triggering conditions for high-risk factors (hard indicators) are that even if the total score is not high, certain extremely dangerous single factors are enough to trigger the warning. The triggering conditions are: any of the following diagnoses, medical histories, or plans appearing in the electronic medical record: a history of VTE (deep vein thrombosis or pulmonary embolism), a confirmed acquired thrombotic tendency (such as protein C / S deficiency, antiphospholipid syndrome, etc.), cancer (the patient is receiving chemotherapy, radiotherapy, or targeted therapy), a major orthopedic surgery plan (such as total hip replacement, total knee replacement, hip fracture surgery), or severe trauma (such as injury, pelvic fracture, or multiple fractures).

[0037] Furthermore, the triggering conditions are dynamically changing. Patient conditions are constantly changing, requiring continuous monitoring by the system. Triggering conditions include: after a patient is transferred to another department (e.g., from the ICU back to a regular ward), the system automatically reassesses, and the new score reaches a high-risk level. 24 hours after surgery, the system automatically re-performs a Caprini assessment, and the score reaches a high-risk level. If a doctor issues a new "absolute bed rest" order, the system recognizes this order and automatically adds a "bed rest" score to the risk assessment, causing the total score to reach the high-risk threshold. If a new cancer diagnosis is confirmed, the system captures this information and recalculates the score after the doctor enters the cancer diagnosis into the medical record. Specific manifestations and examples (workflow) after a warning is triggered: after a warning is triggered, the system does not simply display a pop-up window, but rather initiates a standardized warning and handling process.

[0038] like Figure 8 As shown, step S32 includes: Step S321: If the analysis result is a medium-risk result or a high-risk result, then the anticoagulation strategy is determined in the basic prevention strategy according to the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition and the scoring result; Step S322: If the analysis result is not a medium-risk result or a high-risk result, then the avoidance strategy is determined in the basic prevention strategy based on the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition, and the scoring result.

[0039] Specifically, if the analysis result is a medium-risk or high-risk result, the anticoagulation strategy is determined in the basic prevention strategy based on the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition, and the scoring result (anticoagulation strategy: for patients with low thrombosis risk and high bleeding risk, intermittent pneumatic compression device (IPC) combined with gradient compression stockings is recommended). If the analysis result is not a medium-risk or high-risk result, the avoidance strategy is determined in the basic prevention strategy based on the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition, and the scoring result (avoidance strategy: for patients with a history of peptic ulcer disease, the use of non-steroidal anti-inflammatory drugs (NSAIDs) is automatically excluded; for example, patients with recent intracranial hemorrhage are marked as having absolute contraindications to anticoagulation). The target intervention strategy includes either the anticoagulation strategy or the avoidance strategy.

[0040] For example, such as Figure 9 As shown, upon admission (usually due to high-risk factors such as fractures, surgery, or advanced age), the system immediately performs an automatic assessment. If the VTE risk is extremely high (e.g., Caprini score of 10) and the bleeding risk is also high (e.g., HAS-BLED score of 3), the system automatically generates a personalized VTE prevention plan. This plan is implemented throughout the preoperative and postoperative stages, comprehensively utilizing both pharmacological and mechanical preventative measures. Preoperatively, mechanical prevention is the primary method, such as using a plantar pump and encouraging patients to perform ankle pump exercises. In the critical postoperative stage, pharmacological prevention is initiated in addition to mechanical prevention, prioritizing the use of dapoxetine sodium (2.5 mg subcutaneously once daily), which is relatively safe for renal function, or enoxaparin, which requires dose reduction and close monitoring, as an alternative. Throughout the process, the system simultaneously implements continuous monitoring (including daily physical examination for signs of bleeding and regular blood tests and renal function tests) and patient education (explaining the necessity of prevention and guiding self-observation), thus forming a closed-loop management system covering assessment, intervention, monitoring, and education.

[0041] Furthermore, such as Figure 10 As shown, based on the above-described risk assessment method based on multimodal data, the present invention also provides a risk assessment system based on multimodal data, wherein the risk assessment system based on multimodal data includes: The multimodal data acquisition module 51 is used to acquire the patient's multimodal data, and to match and calculate the multimodal data based on the preset VTE risk assessment rules to obtain the basic risk level and basic prevention strategy. The risk assessment module 52 is used to acquire the patient's dynamic time-series data, adjust the basic risk level based on the dynamic time-series data, generate multiple dynamic risk weights, and fuse all the dynamic risk weights to obtain a scoring result. The intervention strategy execution module 53 is used to acquire the early warning triggering conditions, determine the target intervention strategy in the basic prevention strategy according to the early warning triggering conditions and the scoring results, and execute the target intervention strategy.

[0042] Furthermore, such as Figure 11 As shown, based on the above-mentioned risk assessment method and system based on multimodal data, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 11 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0043] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a risk assessment program 40 based on multimodal data, which can be executed by the processor 10 to implement the risk assessment method based on multimodal data in this application.

[0044] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the risk assessment method based on multimodal data.

[0045] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The terminals communicate with each other via a system bus.

[0046] In one embodiment, when the processor 10 executes the risk assessment program 40 based on multimodal data in the memory 20, the following steps are performed: Acquire patients' multimodal data, and match and calculate the multimodal data based on preset VTE risk assessment rules to obtain basic risk levels and basic prevention strategies; The patient's dynamic time-series data is acquired, the baseline risk level is adjusted based on the dynamic time-series data, multiple dynamic risk weights are generated, and all the dynamic risk weights are fused and calculated to obtain the scoring result. Obtain the early warning triggering conditions, determine the target intervention strategy in the basic prevention strategy based on the early warning triggering conditions and the scoring results, and execute the target intervention strategy.

[0047] The multimodal data includes electronic medical record text and ultrasound reports; The process of acquiring patients' multimodal data, matching and calculating based on preset VTE risk assessment rules, yields a basic risk level and basic prevention strategy, specifically including: The patient's electronic medical record text and ultrasound report are obtained, and multiple assessment values ​​are obtained by matching the electronic medical record text and ultrasound report based on preset VTE risk assessment rules. A basic risk level is calculated based on all the aforementioned assessment values, and a basic prevention strategy is determined based on the basic risk level. The basic risk level includes any one of the following: low basic level, second-lowest basic level, intermediate basic level, and high basic level.

[0048] The dynamic time-series data includes surgery type, blood loss, D-dimer time-series data, and vital signs data. The process of acquiring dynamic time-series data of patients, adjusting the baseline risk level based on the dynamic time-series data, generating multiple dynamic risk weights, and fusing all the dynamic risk weights to obtain a scoring result specifically includes: Obtain the patient's surgical type, blood loss, D-dimer time-series data, and vital signs; The baseline risk level is adjusted based on the surgical type, blood loss, D-dimer time series data, and vital signs to obtain multiple risk weight values. All risk weight values ​​are then fused together to obtain a scoring result.

[0049] Specifically, the adjustment of the basic risk level based on the surgical type, blood loss, D-dimer time-series data, and vital signs yields multiple risk weight values. All risk weight values ​​are then fused and calculated to obtain a scoring result, which includes: The surgical type, the amount of blood loss, the D-dimer time series data, and the vital signs are input into a deep learning model for analysis to obtain the analysis results. Based on the analysis results, any one of the low-level basic level, the second-lower-level basic level, the intermediate-level basic level, and the high-level basic level is adjusted to obtain multiple risk weight values. All risk weight values ​​are then integrated and calculated to obtain the scoring result.

[0050] The analysis results include any one of low-risk, medium-risk, and high-risk results; The process involves inputting the surgical type, blood loss, D-dimer time-series data, and vital signs into a deep learning model for analysis to obtain the analysis results, specifically including: The surgical type, the amount of blood loss, the D-dimer time series data, and the vital signs are divided to obtain a training set and a test set. Based on the initial deep learning model, the LSTM network of the initial deep learning model is trained according to the training set to obtain the deep learning model. The test set is input into a deep learning model for analysis to obtain any one of the following results: low risk, medium risk, or high risk.

[0051] The early warning triggering conditions include core triggering conditions, high-risk factor triggering conditions, and dynamic change triggering conditions; The process of obtaining early warning trigger conditions, determining a target intervention strategy within the basic prevention strategy based on the early warning trigger conditions and the scoring results, and executing the target intervention strategy specifically includes: Obtain the current stage of diagnosis and treatment, along with the core triggering conditions, the high-risk factor triggering conditions, and the dynamic change triggering conditions; Based on the core triggering conditions, the high-risk factor triggering conditions, the dynamic change triggering conditions, and the scoring results, a target intervention strategy is determined in the basic prevention strategy, and the target intervention strategy is executed.

[0052] The target intervention strategy includes either an anticoagulation strategy or an avoidance strategy. The step of determining a target intervention strategy within the basic prevention strategy based on the core triggering conditions, the high-risk factor triggering conditions, the dynamic change triggering conditions, and the scoring results, and then executing the target intervention strategy, specifically includes: If the analysis result is a medium-risk or high-risk result, then the anticoagulation strategy is determined in the basic prevention strategy based on the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition, and the scoring result. If the analysis result is not a medium-risk or high-risk result, then the avoidance strategy is determined in the basic prevention strategy based on the core triggering condition, the high-risk factor triggering condition, the dynamic change triggering condition, and the scoring result.

[0053] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a risk assessment program based on multimodal data, and the risk assessment program based on multimodal data, when executed by a processor, implements the steps of the risk assessment method based on multimodal data as described above.

[0054] In summary, this invention provides a risk assessment method, system, terminal, and storage medium based on multimodal data. The method includes: acquiring multimodal data of a patient; matching and calculating the multimodal data based on preset VTE risk assessment rules to obtain a basic risk level and a basic prevention strategy; acquiring dynamic time-series data of the patient; adjusting the basic risk level based on the dynamic time-series data to generate multiple dynamic risk weights; fusing all the dynamic risk weights to obtain a scoring result; acquiring early warning triggering conditions; determining a target intervention strategy in the basic prevention strategy based on the early warning triggering conditions and the scoring result; and executing the target intervention strategy. This invention iteratively optimizes patient VTE risk through a dynamic weight fusion mechanism, significantly improving the accuracy of risk assessment and intervention strategies.

[0055] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal system that includes that element.

[0056] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0057] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for risk assessment based on multi-modal data, characterized in that, The risk assessment method based on multi-modal data comprises: Obtaining multi-modal data of a patient, matching and calculating the multi-modal data based on a preset VTE risk assessment rule to obtain a basic risk level and a basic prevention strategy; Obtaining dynamic time series data of the patient, adjusting the basic risk level based on the dynamic time series data to generate a plurality of dynamic risk weights, and fusing and calculating all the dynamic risk weights to obtain a score result; Obtaining a warning trigger condition, determining a target intervention strategy in the basic prevention strategy based on the warning trigger condition and the score result, and executing the target intervention strategy.

2. The method for risk assessment based on multi-modal data according to claim 1, wherein, The multi-modal data comprises electronic medical record texts and ultrasound reports; The method for obtaining multi-modal data of a patient, matching and calculating the multi-modal data based on a preset VTE risk assessment rule to obtain a basic risk level and a basic prevention strategy comprises: Obtaining the electronic medical record texts and the ultrasound reports of the patient, matching the electronic medical record texts and the ultrasound reports based on a preset VTE risk assessment rule to obtain a plurality of evaluation values; Calculating all the evaluation values to obtain a basic risk level, and determining a basic prevention strategy based on the basic risk level; The basic risk level comprises any one of a low basic level, a sub-low basic level, a medium basic level and a high basic level.

3. The method for risk assessment based on multi-modal data according to claim 2, characterized in that, The dynamic time series data comprises a surgery type, a bleeding amount, D-dimer time series data and vital sign data; The method for obtaining dynamic time series data of a patient, adjusting the basic risk level based on the dynamic time series data to generate a plurality of dynamic risk weights, and fusing and calculating all the dynamic risk weights to obtain a score result comprises: Obtaining the surgery type, the bleeding amount, the D-dimer time series data and the vital signs of the patient; Adjusting the basic risk level based on the surgery type, the bleeding amount, the D-dimer time series data and the vital signs to obtain a plurality of risk weight values, and fusing and calculating all the risk weight values to obtain a score result.

4. The method for risk assessment based on multi-modal data according to claim 3, characterized in that, The method for adjusting the basic risk level based on the surgery type, the bleeding amount, the D-dimer time series data and the vital signs to obtain a plurality of risk weight values, and fusing and calculating all the risk weight values to obtain a score result comprises: Inputting the surgery type, the bleeding amount, the D-dimer time series data and the vital signs into a deep learning model for analysis to obtain an analysis result; Adjusting any one of the low basic level, the sub-low basic level, the medium basic level and the high basic level based on the analysis result to obtain a plurality of risk weight values, and fusing and calculating all the risk weight values to obtain a score result.

5. The method for risk assessment based on multi-modal data according to claim 4, characterized in that, The analysis result comprises any one of a low-risk result, a medium-risk result and a high-risk result. The operation type, the amount of bleeding, the D-dimer time series data and the vital signs are input into a deep learning model for analysis to obtain an analysis result, specifically comprising: The operation type, the amount of bleeding, the D-dimer time series data and the vital signs are divided to obtain a training set and a test set, and based on an initial deep learning model, the LSTM network of the initial deep learning model is trained according to the training set to obtain a deep learning model; The test set is input into the deep learning model for analysis to obtain any one of a low-risk result, a medium-risk result and a high-risk result.

6. The method of risk assessment based on multi-modal data according to claim 5, wherein, The pre-warning trigger condition includes a core trigger condition, a high-risk factor trigger condition and a dynamic change trigger condition; The pre-warning trigger condition is obtained, a target intervention strategy is determined in the basic prevention strategy according to the pre-warning trigger condition and the score result, and the target intervention strategy is executed, specifically comprising: The current diagnosis and treatment stage and the core trigger condition, the high-risk factor trigger condition and the dynamic change trigger condition are obtained; The target intervention strategy is determined in the basic prevention strategy according to the core trigger condition, the high-risk factor trigger condition, the dynamic change trigger condition and the score result, and the target intervention strategy is executed.

7. The method of risk assessment based on multi-modal data according to claim 6, characterized in that, The target intervention strategy includes any one of an anticoagulation strategy and an avoidance strategy; The target intervention strategy is determined in the basic prevention strategy according to the core trigger condition, the high-risk factor trigger condition, the dynamic change trigger condition and the score result, and the target intervention strategy is executed, specifically comprising: If the analysis result is a medium-risk result or a high-risk result, the anticoagulation strategy is determined in the basic prevention strategy according to the core trigger condition, the high-risk factor trigger condition, the dynamic change trigger condition and the score result; If the analysis result is not a medium-risk result or a high-risk result, the avoidance strategy is determined in the basic prevention strategy according to the core trigger condition, the high-risk factor trigger condition, the dynamic change trigger condition and the score result.

8. A risk assessment system based on multi-modal data, characterized in that, The risk assessment system based on multi-modal data comprises: A multi-modal data acquisition module is configured to acquire multi-modal data of a patient, match and calculate the multi-modal data based on a preset VTE risk assessment rule, obtain a basic risk level and a basic prevention strategy; A risk assessment module is configured to acquire dynamic time series data of a patient, adjust the basic risk level according to the dynamic time series data, generate a plurality of dynamic risk weights, and fuse and calculate all the dynamic risk weights to obtain a score result; An intervention strategy execution module is configured to acquire a pre-warning trigger condition, determine a target intervention strategy in the basic prevention strategy according to the pre-warning trigger condition and the score result, and execute the target intervention strategy.

9. A terminal, characterized by comprising: The terminal comprises a memory, a processor, and a multi-modal data based risk assessment program stored on the memory and executable on the processor, the multi-modal data based risk assessment program, when executed by the processor, implements the steps of the multi-modal data based risk assessment method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a multi-modal data based risk assessment program, the multi-modal data based risk assessment program, when executed by a processor, implements the steps of the multi-modal data based risk assessment method according to any one of claims 1-7.

Images